
Photoacoustic Remote Sensing Peak Detector
Photomedicine Labs
The purpose of this peak detector was to both increase the accuracy
and decrease the magnitude of data collection in photoacoustic
remote sensing (PARS) systems. Currently, the peaks of each pulse in
the time domain obtained by taking 64 samples and iterating
through each to find the greatest magnitude. The difficulty lies in
the miniscule pulse wavelength of 20ns, as well as the 50kHz
frequency that they are produced at. The design sought to use an
OPA615 transconductance amplifier due to its minimal capacitance
and high speed. Furthermore, Texas Instruments (the manufacturer)
provides a basic layout for a peak detection circuit that could be
adapted to this high precision application.
The first step in the designing this system was creating a simulation
to gain a better understanding of what components performed at the
level necessary to hold the peaks. LTSpice was utilized as many
companies provide public simulation models of their components,
making it an effective way of comparing different diode and
MOSFET performances. The next step was to create a breadboard,
and subsequent perfboard prototype. The parasitic capacitance in
the breadboard itself meant that the system was unable to capture
small peaks but could easily hold longer pulses and be reset at the
required 50kHz frequency. The perfboard prototype was much more
effective, and despite the great amounts of noise present, it was able
to prove that the system was capable of capturing peaks of pulses
with the same wavelength as the PARS signals.
Due to the high-speed nature of the system, the PCB was designed to
be as efficient as possible. Certain design features, such as internal
unbroken ground planes, short and straight signal paths,
decoupling capacitors on all power supplies, and impedance
matching the traces to match the BNC connectors were all crucial to
ensure signal accuracy and clarity. The three most effective diodes
(determined from the simulations) were selected, and two different
boards were designed to have both 1 and 2 diode configurations.
This meant 6 total boards were manufactured and tested. Subsequent
comparisons between the accuracy of their outputs allowed for a
definitive choice of the most effective design.
The first lesson I learned while working on this project was the
importance of accurate simulation. At first, I used the basic diode
and NMOS models in the LTSpice simulations, which provided
excellent results. Upon building an initial breadboard prototype
using parts I had lying around, I was surprised to find that it was
unable to detect pulses with wavelengths an entire order of
magnitude greater than what was desired. This made it clear that
accurate simulation is imperative, as using the idealized components
provided by LTSpice was essentially useless in determining the
system’s efficacy. Another big takeaway from this project was the
importance of prototyping for proof of concept. The diodes used for
the PCB all had a footprint of just 0.3x0.6mm, meaning they were
simply impossible to solder by hand. Unfortunately, no larger diodes
were able to match their performance in simulation. To get around
this, a diode that was larger and provided much worse component
was used, as a proof of concept was all that was necessary. To look at
the final board design, please look here.
Key Takeaways and Future Considerations